Pre-launch· Token not live yet — no rewards or treasury until launchWhat this means
← Model
Decision trees — retrained on new data

TRAIN v0.13

Same configuration as v0.04 (best so far on the headline benchmark), retrained on the newest dataset.
Completed

Definition

Model type
Decision-tree ensemble
Signals used
Market signals
Technical details
max_depth
2
subsample
0.8
n_estimators
120
learning_rate
0.05
min_samples_leaf
10
Random seed
1013
Config hash
3e4ab42c45fd72c2
Created
05 Oct 11:49
Completed
05 Oct 11:49
Artifact
models/v0.13/model.json
sha256 4548ba75de19f602…

Dataset & compute

Dataset
Solana Launch Dataset v9
3,937 snapshots · 789 launches · sha256 78a0a22f0364
Training runs
1
Compute
TRAIN training server (CPU)
GPU hours
—
no GPU used
CPU time
8.3 s
Provider cost
$0.00
nothing purchased

Benchmarks

held-out test set · change vs v0.12
Spotting crashes (−70% within 1h)
0.946
Spotting survivors (still traded after 1h)
0.948
Spotting 2× runs (within 1h)
0.830
Calling the 1h direction (down / flat / up)
52.6%
Spotting creator dumps (within 1h)
0.835
Spotting crashes (within 6h)
insufficient data
Spotting survivors (6h)
insufficient data
Calling the 6h direction
insufficient data
Spotting survivors (24h)
insufficient data
Spotting migrations (within 24h)
insufficient data

Validation metrics (during training)

collapse_1h
AUC 0.992
n=435
collapse_6h
—
n=0
reach_2x_1h
AUC 0.971
n=435
survival_1h
AUC 0.914
n=435
survival_6h
—
n=0
survival_24h
—
n=0
migration_24h
—
n=0
trajectory_1h
acc 96.6%
n=435
trajectory_6h
—
n=0
creator_exit_1h
AUC 0.799
n=131

Validation launches are separate from training launches and from the locked test set. Benchmarks above are the public numbers.

What this version relies on

held-out validation launches
This version was trained before TRAIN started measuring which signals each model relies on.

For each signal, how much the score drops when that signal is scrambled across launches. TRAIN learns these weights itself from the data; nobody hand-codes them.

Training runs

Attempt 1Succeeded
05 Oct 11:49 · 00:00:04 · 3,502 examples · $0.00

Timeline

  1. 05 Oct 11:49
    TRAIN v0.13 not deployed; v0.09 stays live
    Reason: average score across 5 tests 0.800 vs 0.812 for the live v0.09.
  2. 05 Oct 11:49
    Benchmark completed: TRAIN v0.13
    Crash-spotting score improved 0.911 → 0.918 vs live v0.09.
  3. 05 Oct 11:49
    Benchmark started: TRAIN v0.13
  4. 05 Oct 11:49
    Training completed: TRAIN v0.13
    00:00:04 · trained on 3502 examples · $0 compute cost
  5. 05 Oct 11:49
    Training started: TRAIN v0.13
    Solana Launch Dataset v9 · decision-tree ensemble · TRAIN training server
  6. 05 Oct 11:49
    TRAIN v0.13 queued
    Same configuration as v0.04 (best so far on the headline benchmark), retrained on the newest dataset.